An adaptive correction technique for DGPS using Recurrent Wavelet Neural Network

Systematic errors in Global Positioning System (GPS) were related to the atmosphere, imprecise orbit, satellite distribution geometry, multipath, satellite and receiver clock, and Selective Availability (SA). Using DGPS corrections prediction can reduce these errors to a certain extent, with the exception of SA. This makes it possible to enhance the accuracy of GPS positioning by a factor of twenty five. This paper presents a Recurrent Wavelet Neural Network (RWNN) for improving positioning accuracy. Method validity is verified with experimental data from an actual data collection, before and after SA. The results show very clearly the dramatic change in the error value. It also shows that by using proposed method over a long period of time, 0.7m GPS accuracy can be achieved. Also, it is shown that RWNN outperforms single WNN and RNN.